Subset assignment when only 'i' is provided

[joshuaulrich]

There's slightly different behavior between "matrix" and xts/zoo when it comes to subset and subset-assignment calls that only specify i. For example, using the data from a stack overflow question:

dDate <- as.Date("1991-01-02")
m1 <- cbind(trade_level = 1, t2 = 3)
m2 <- cbind(trade_level = 5, t2 = 0.053)
x1 <- xts(m1, dDate)
x2 <- xts(m2, dDate)
z1 <- zoo(m1, dDate)
z2 <- zoo(m2, dDate)

Consider the behavior of each class if only i is specified when subsetting. The matrix object returns the first element, while both the xts and zoo objects return all the columns for the first row.

m1[1L]
# [1] 1
x1[1L]
#            trade_level t2
# 1991-01-02           1  3
z1[1L]
#            trade_level t2
# 1991-01-02           1  3

Now compare the 3 classes' behavior when assigning a new row to a subset when only i is provided:

m1[1L] <- m2
# Warning message:
# In m1[1L] <- m2 :
#   number of items to replace is not a multiple of replacement length
x1[1L] <- x2
# Warning message:
# In x1[1L] <- x2 :
#   number of items to replace is not a multiple of replacement length
z1[1L] <- z2
# Warning message:
# In z1[1L] <- z2 :
#   number of items to replace is not a multiple of replacement length

They all warn, and none of the subset assignment calls alter the original object. The result isn't too surprising for matrix, since the subset call only returns one element. But it's a bit surprising for xts and zoo because the subset call returns the entire row, so it seems like you should be able to replace an entire row.